6 min read · 12 sources

Bitget’s $388 million exploit is the largest crypto theft this year, and the exchange is covering all losses from its 5,500 BTC User Protection Fund.

Bitget Starts Phased Withdrawal Resumption After $388 Million Exploit

Source: theblock.co ↗

Bitget began restoring withdrawals in stages on September 29, following an exploit on September 24 that drained roughly $388 million from its hot and warm wallets. That makes it the largest reported crypto theft this year. BTC withdrawals resumed at 8 AM UTC on September 29, with ETH following the same day, USDT on September 30, and all remaining assets, fiat, and P2P by October 2.

Private keys, cold wallets, and user balances were not affected. Bitget says losses will be fully covered by its 5,500 BTC User Protection Fund, and it suspects the attackers were “sophisticated” and “state-backed,” pointing at North Korea. For engineers running exchanges or custody systems, the lesson is the usual one: hot and warm wallet exposure is the attack surface, and a third-party vulnerability - not key compromise - was the entry point.

Strategy Proposes Daily Dividends to Push STRC Toward $100

Source: coindesk.com ↗

Strategy is asking shareholders to approve daily dividend accrual for its four U.S.-listed preferred stocks - STRF, STRC, STRK, and STRD - with a vote on October 15. The proposal changes payment timing only; dividend rates and total payouts stay the same. Daily payments, settled each subsequent business day, are designed to attract income investors and dampen price swings around ex-dividend dates.

The move targets the persistent discount on STRC, which has traded below its $100 liquidation preference. Engineers should note that daily accrual smooths the price decay between payments, but the real value driver remains Strategy’s 846,000 BTC holdings, with an average cost of $75,416. The dividend mechanics are a distraction; the balance sheet is the trade.

The Cryptographic World Computer

Source: vitalik.eth.limo ↗

Vitalik Buterin’s latest post frames Ethereum’s trajectory as a shift from “blockchain” to “cryptographic world computer” - a hybrid architecture mixing traditional consensus with modern cryptography. Key changes include verification moving from full re-execution to PeerDAS sampling with SNARK verification, consensus shifting from PoW to optimized PoS, and block construction moving from single miners to multi-party systems. The post compares the 2009 Bitcoin whitepaper section-by-section against Ethereum in 2015, 2025, and 2030.

Buterin argues that modern networks merge Satoshian ideas with 50 years of cryptographic research, including obfuscation as a longer-term primitive. For engineers, this is a fundamental redesign of trust and scalability assumptions: tunable computation, ZK-proofs at the base layer, and multi-party block construction. The “world computer” label is not marketing; it’s a claim that the architecture can support properties no pure blockchain can.

Solana Tests Alpenglow: Finality from 12.8 Seconds to 150 Milliseconds

Source: coindesk.com ↗

Solana started testing Alpenglow, an upgrade that replaces TowerBFT consensus with Votor, where validators finalize blocks after one or two rounds of direct voting. The change eliminates on-chain vote accumulation across 32 slots, replacing it with direct validator-to-validator messaging that settles in one or two rounds - no changes to wallets or user-facing transaction flows. The testnet migration requires Agave 4.3, and Firedancer and Frankendancer clients do not yet support the upgrade. A tentative mainnet date of September 28 has not been confirmed.

The finality reduction from 12.8 seconds to 150 milliseconds would cut the hold periods exchanges impose before crediting deposits and the wait bridges require before releasing cross-chain funds. But the upgrade’s success depends on validator coordination and software readiness. Faster finality makes Solana more viable for exchanges, bridges, and merchants, but only if the validators actually run the new client.

EIP-1559's Burn Mechanism and the Economic Abstraction Problem

Source: threadreaderapp.com ↗

A thread on EIP-1559 explains that the burn mechanism captures economic value and removes it from supply, preventing “economic abstraction” where other assets could pay protocol fees. Without the burn, the argument goes, the network’s native token loses its role as the fee currency, and the economic security model breaks down. The thread frames this as a “bulletin board, not a world computer” problem: permissionless actions like liquidations depend on the base layer’s fee market staying intact.

For engineers, this is about incentive design. The burn is not just deflationary; it’s what keeps ETH as the unit of account for gas, which in turn secures the chain. Remove it, and you get a race to the bottom on fee tokens.

Why AI Agents Could Make Bank Runs Faster and More Violent

Source: threadreaderapp.com ↗

A separate thread argues that DeFi TVL at $42 billion, a 2.5-year low, is not necessarily bearish. The real story isn’t an imminent agentic bank run but the collapse of banks and brokers like Schwab and IBKR that rely on idle sweep-account balances. When agentic bank runs do happen, they’ll be faster and more violent than human ones like SVB’s, because AI agents have much lower behavioral diversity than humans.

The thread also critiques Ledger’s Secure Element, clarifying it is not like Apple’s Secure Enclave - keys can sign but not leave the device, contrary to initial assumptions. For engineers, the TVL analysis is useful for market positioning, and the Secure Element distinction matters for hardware wallet security assessments. The bank run thesis is speculative, but the mechanics are sound: agents react in milliseconds, not days.

Hegseth Holds Bitcoin in a Coinbase Wallet, Filing Shows

Source: decrypt.co ↗

Defense Secretary Pete Hegseth’s filing shows no Bitcoin trades above $1,000, with bank cash accounts over $1 million and Lockheed and Northrop stakes in the $1,001 - $15,000 range. Two mortgages on a personal residence are each in the $1,000,001 - $5,000,000 range, taken out in 2022 and 2025. The disclosure uses ranges, which complicates total valuation, but the lack of trades suggests a long-term hold.

This is notable compared to President Trump’s June filing, which reported over $50 million in Bitcoin and $1.2 billion in crypto-related earnings. Hegseth’s holding is small by comparison, but it establishes Bitcoin ownership at the cabinet level, with Coinbase as the custodian. For engineers, the ethical implications of a defense secretary holding crypto are murky, but the disclosure is transparent.

Circle's Tazapay Acquisition: The Off-Ramp Is the Business

Source: threadreaderapp.com ↗

Circle’s acquisition of Tazapay is primarily about securing an off-ramp for USDC, not just B2B payments. Tazapay claims $25 billion in annualized volume, with on-chain redemptions to Circle rising from $190 million in June to $562 million in the first 25 days of September, averaging $300,000 per transfer. Each transfer involves three steps: client payment, contract forwarding to treasury, and a 2 basis point fee.

September’s growth is entirely Ethereum-based, nearly tripling month-over-month, indicating a large new client. The real business is the FX spread at the off-ramp, not the blockchain layer itself - Tazapay’s product is one-directional (stablecoin in, fiat out) with almost no on-chain revenue. For engineers, this highlights the strategic importance of fiat rails for stablecoin liquidity. Circle isn’t buying a payment processor; it’s buying a channel that already burns over half a billion dollars in USDC per month.

Quick Links

Source: threadreaderapp.com ↗

Plasma One offers strong USD-EUR transfer rates: sending $1,015 gives recipients €890.45, only about 0.14% below the mid-market rate. Binance Wallet adds gas-fee payment in USDT across four chains, so users no longer need to hold small amounts of ETH, SOL, BNB, or TRX just to move USDT. Vitalik’s “Snowmoon” post outlines a legal theory, deemed plausible by Kimi K3, that allows others to turn the work into a movie or vibe-coded anime, provided they open-source the pipeline (AI prompts, scripts, task-specific models). A thread on crypto builder virality critiques the current landscape, contrasting it with pre-algorithmic virality, but the content is truncated.